Malicious Claude Code Extension Update Hit Up to 5,000 Developers, Cursor Says

A bad version of the Claude Code extension spent about eleven hours in editor marketplaces yesterday, and the open question is whose credentials published it. Cursor's notice says Anthropic didn't write it; the headline says Anthropic was hacked; the people in the comment threads are working out which of those is closer. Cursor's security notice names version 2.1.313, a window of roughly 2:27 a.m. to 1:35 p.m. Pacific on October 29, an SSH key labeled svc-telemetry, and a cap of 5,000 affected users. It also warns that uninstalling may not be enough. Double Pulsar carries the 'Anthropic hacked' headline, and Kevin Beaumont's Mastodon post reports backdoors in official Claude Code extensions. Both reached Hacker News this morning, where commenters raised the question of who owns the publisher account and noted that Open VSX has no extension signing. CrowdStrike via NBC News says China-linked hackers used an AI tool called ARTEX with a DeepSeek model against more than 200 organizations in 11 days, and Wired reports an Accenture employee's personal Claude agent ran on a Defense Department network for two months. Together they show agents arriving inside networks without anyone deciding they should. OpenAI's post on detecting misalignment, a Reddit summary of an Anthropic study in which analysts followed an agent's hidden lean 86 percent of the time, and Forbes on who can shut an agent down cover monitoring on the model side and deference on the human side. Bloomberg via Techmeme reports Anthropic as 34.5 percent of SpaceX's third-quarter revenue; the WSJ on the S-1 covers the TSMC dependency; The Information has Nvidia in talks on credit guarantees for Nscale; and the FT visits Nscale's West Virginia campus. TechCrunch and SiliconANGLE on Mistral Large 4: 1 trillion parameters, 52 billion active, pitched at cyber defense, with no license named in either piece. The Next Web also covers the release. The FT on Alphabet's 12 percent drop, Bloomberg on Microsoft's bond sale, TechCrunch on Jassy, the AI Daily Brief recap and CNBC on Foxconn show the same spending read three different ways. In the Brief: DeepSeek's round, the spy-plot report and Nvidia's response; Axios on Reddit going logged-in, its Techmeme entry and the CMA's Meta probe; Wired on &Truth and Sonar's talk; Vals AI's talk and Decagon's post; and AWS on MoEngage, SiliconANGLE on access control and Serval's raise.